AI Agent for Automated Product Data Enrichment from Images

Unlock Automated Product Cataloging with this AI Agent

This powerful n8n workflow acts as an AI Agent dedicated to enriching your product data. It starts by fetching items from your Airtable base that have product images but lack detailed information. The agent then performs a multi-step analysis: first, using OpenAI's vision capabilities to understand the image content. Then, an advanced AI (GPT-4o based Langchain agent) equipped with web research tools (reverse image search via SerpAPI and webpage scraping via Firecrawl) gathers precise product attributes like title, description, model, material, color, and condition. Finally, it structures this information and updates your Airtable, transforming raw images into comprehensive product listings.

This agent showcases advanced n8n capabilities, including self-triggering workflows to manage tool execution for the AI agent, providing a robust framework for complex AI-driven tasks.

Key Features & Benefits

• Automated Image Analysis: Leverages OpenAI's vision models (specifically GPT-4o with vision) to interpret product images and extract initial details.
• Intelligent Web Research: Employs an AI agent that can use tools for:
  • Reverse Image Search: Utilizes SerpAPI to find similar images and product sources online.
  • Web Scraping: Uses Firecrawl to extract relevant text content from identified web pages.
• Advanced AI Reasoning: Powered by OpenAI's GPT-4o via a Langchain agent, enabling complex decision-making and tool orchestration.
• Structured Data Extraction: Ensures consistent and accurate product attributes (title, description, model, material, color, condition) using a structured output parser.
• Seamless Airtable Integration: Reads unprocessed items from Airtable and writes back enriched data automatically.
• Boost Efficiency: Drastically reduces manual data entry for product cataloging and enrichment.
• Improve Data Quality: Enhances product listings with accurate, detailed, and consistent information.
• Scalable Automation: Handles large volumes of products, ideal for growing e-commerce businesses or extensive inventories.

• An n8n instance (Cloud or self-hosted).
• OpenAI API Key with access to a vision-capable model (e.g., GPT-4o).
• Airtable account and Personal Access Token.
• Airtable Base set up with at least the following fields: 'Image' (Attachment type for product photos), 'AI_status' (Checkbox/Boolean, initially false), and fields for 'Title', 'Description', 'Model', 'Material', 'Color', 'Condition' (Text or Single Select types).
• SerpAPI API Key (for the reverse image search tool).
• Firecrawl API Key (for the web scraping tool).

1. Download the n8n workflow JSON file.
2. Import the workflow into your n8n instance.
3. Configure Credentials:
   • Airtable: In n8n, go to Credentials -> New. Search for Airtable, select 'Airtable Personal Access Token', and add your token. Select this credential in the 'Get Applicable Rows' and 'Enrich Product Rows' nodes.
   • OpenAI: In n8n, go to Credentials -> New. Search for OpenAI, select 'OpenAI API', and add your API key. Select this credential in the 'OpenAI Chat Model1' (for the agent) and 'Analyse Image' (for vision analysis) nodes.
   • SerpAPI: In n8n, go to Credentials -> New. Search for SerpApi, select 'SerpApi API', and add your API key. Select this credential in the 'SERP Google Reverse Image API' node.
   • Firecrawl: In n8n, go to Credentials -> New. Select 'Header Auth'. Name it (e.g., 'Firecrawl API Key'). For 'Header Name', enter Authorization. For 'Header Value', enter Bearer YOUR_FIRECRAWL_API_KEY (replace with your actual key). Select this credential in the 'Firecrawl Scrape API' node.
4. Configure Airtable Nodes:
   • 'Get Applicable Rows' node: Enter your Airtable Base ID and Table ID. Verify the filterByFormula targets records with an image and AI_status=FALSE().
   • 'Enrich Product Rows' node: Enter your Airtable Base ID and Table ID. Ensure the id field for matching is correctly mapped from {{ $('Get Applicable Rows').item.json.id }}. Map the output fields (e.g., {{ $json.output.output.color }}) to your corresponding Airtable column names.
5. Review AI Agent Configuration:
   • 'Analyse Image' node: Check the prompt to ensure it aligns with your initial image analysis needs.
   • 'Object Identifier Agent' node: Review the system prompt. You can customize it to change the agent's persona, instructions, or how it uses tools.
   • 'Structured Output Parser' node: The JSON schema defines the expected output. Adjust if your required attributes differ.
6. Test the Workflow: Manually trigger the workflow ("When clicking 'Test workflow'"). Ensure you have a test record in Airtable that meets the criteria in 'Get Applicable Rows'.
7. Activate for Automation: Once testing is successful, activate the workflow. For continuous operation, replace the manual trigger with an 'Airtable Trigger' (e.g., 'On New Record Matching Criteria') or a 'Schedule Trigger'.